Pulmonary artery hypertension (PAH), which is diagnosed by an abnormal increase of blood pressure in the pulmonary artery, can be a severe disease, leading to heart failure. In recent years, medical imaging, such as echocardiography, magnetic resonance imaging (MRI), and computed tomography (CT), has been widely used due to its non-invasive property. Right pulmonary artery (RPA) wall distensibility derived from CT angiography was reported to serve as a reliabile marker for the diagnosis of PAH. This study presented a robust method for automatic segmentation of artery based on CT angiography. The algorithm can be divided into two steps: generation of initial contour and refinement of edge. In the first step, a series of original images at different cardiac phases were thresholded to retrieve appropriate intensity window of vessels, followed by the determination of initial contours by a series of morphological image processing on the binary images with two simple manual initializations. Initial contours without touching can be taken as the final results of segmentation, when others need further refinement of edge. In the second step, the center of vessel was automatically located by an ellipse fitting method and then the ray casting algorithm was applied to search for possible edge. Disconnected segments of edge will be linked to complete the vessel segmentation. Furthermore, cross-sectional areas of arteries at different cardiac phases can be measured and used to obtain distensibility. In this study, artery wall distensibility of patients and healthy subjects was evaluated on four vessels, including aorta, main pulmonary artery, right and left pulmonary artery. In addition, segmentation results of five subjects were compared with those obtained by manual selection to evaluate the reliability of the proposed method.